Skip to main content

AI Is Not a Technology Decision

GPT-4 just launched and the boardroom conversation is about models. It should be about business models.
20 March 2023·4 min read
Isaac Rolfe
Isaac Rolfe
Managing Director
enterprise-aistrategyleadership
GPT-4 launched last week. It's multimodal. It's smarter. It passes the bar exam. And it's almost entirely irrelevant to your enterprise AI strategy.
Here's the pattern I'm seeing in boardrooms across New Zealand right now:
  1. Executive reads about GPT-4 in the weekend papers
  2. Monday morning email: "What's our AI strategy?"
  3. CTO gets tasked with "doing something with AI"
  4. Three months later, a proof-of-concept nobody uses
The problem isn't the technology. The problem is that the entire conversation starts in the wrong place.
AI is not a technology decision. It's a business design decision.
When you frame AI as a technology problem, you get technology answers: model selection, infrastructure, API costs, compute requirements. All important. All secondary.
When you frame AI as a business design problem, you get different questions entirely:
  • Which of our processes create the most value but consume the most time?
  • Where does knowledge get stuck? In people's heads, in legacy systems, in email chains?
  • What would our operation look like if our best person's judgement was available at every decision point?
  • How do we measure whether AI is working? Not "is it accurate?" but "is the business better?"
These are strategy questions, not technology questions. And they need to be answered before anyone opens a terminal.
The GPT-4 trap is the same trap every major technology shift creates. The internet wasn't about HTML. It was about new business models. Mobile wasn't about apps. It was about location-aware, always-connected customers. Cloud wasn't about servers. It was about operational flexibility.
AI isn't about models. It's about redesigning how your organisation creates and delivers value.
The companies that treated the internet as a "technology decision" built brochure websites. The companies that treated it as a business decision built Amazon.
The same split is happening right now with AI. One group is asking "which model should we use?" The other is asking "how should we redesign our claims operation?" or "what would our advisory practice look like if every consultant had access to our collective knowledge?"
My advice: put the technology conversation on hold for 30 days. Spend that time mapping where your organisation's knowledge lives, where decisions get bottlenecked, and where your best people spend time on work that doesn't require their expertise.
Then, and only then, ask how AI might help.
The model you use will matter eventually. But it matters far less than the clarity of the problem you're solving and the quality of the data you're working with.
GPT-4 is impressive. GPT-5 will be more impressive. The enterprise that waits for the perfect model will be waiting forever. The enterprise that starts with the right question will be compounding value long before the hype cycle settles.